reasoning-trace-optimizer

Debug AI agent failures by analyzing interleaved thinking traces to expose decision points and tool usage. This skill diagnoses errors by evaluating reasoning steps between tool calls, providing concrete evidence of failures like context degradation or instruction drift.

Interleaved thinking analysis reviews the reasoning generated between tool calls to identify failure modes. It enables prompt optimization by detecting issues like tool confusion and instruction drift, generating targeted prompt improvements to validate in iterative runs.

Detect context degradation and tool confusion through automated pattern detection on reasoning traces. The analysis identifies these failure modes across AI agent stacks and provides actionable recommendations with concrete evidence to resolve them.

Yes, the optimizer analyzes MiniMax M2.1 interleaved thinking traces. It applies to debugging sessions and reasoning-trace analysis across AI agent stacks, enforcing structured evaluation to generate trace-based quality improvements.

Generate targeted prompt improvements by analyzing reasoning traces to diagnose the exact failure points. The skill integrates prompt optimization by producing actionable recommendations and validating them through iterative agent runs to reduce errors.

Instruction drift occurs when an agent deviates from its original directives during multi-step tool usage. Analyzing interleaved thinking traces reveals these decision points, allowing you to apply structured evaluation and generate corrective prompt improvements.